[PATCHES] A module inliner pass with a greedy call site queue

James Molloy james at jamesmolloy.co.uk
Wed Aug 6 02:07:19 PDT 2014


Hi Chandler,

Thanks for the clarification. So this new GreedyInliner just adds global
context awareness to the current inliner? That sounds like a win to me. But
again, I'm not an inliner expert :)

Cheers,

James


On 6 August 2014 09:46, Chandler Carruth <chandlerc at google.com> wrote:

>
> On Wed, Aug 6, 2014 at 1:11 AM, James Molloy <james at jamesmolloy.co.uk>
> wrote:
>
>> Hi Nick,
>>
>> I'm not an expert on inlining algorithms so please excuse my naivite. But
>> usually these "top-down versus bottom-up" arguments (in other domains, at
>> least), come to the obvious conclusion that both have merits so let's
>> create a hybrid. Why is this not the case here too?
>>
>
> FYI, the current inliner *is* a hybrid. It does deferred top-down inlining
> when it can do so safely during its bottom-up walk.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140806/d7015a4a/attachment.html>


More information about the llvm-commits mailing list